Everything under Cross Attention, SDNQ Attention and Attention Dispatcher
shipped with no hint, which left the layering invisible: the SDP kernel boxes
are candidates torch chooses from per call rather than a selection, and the
Flash box is torch's own build of the kernel rather than the flash-attn package
that the Flash attention override installs.
- cover attention method, sdp kernels, sdp overrides and attention slicing
- give the dynamic attention rates their unit and the estimate they compare against
- cover the sdnq attention kernel settings, including the head dimension clamp
that the hadamard group size resolves through
- cover the diffusers attention dispatcher and the backend names it takes
- record the constraints each override serves, leaving the throughput comparison
to the workload rather than naming a winner
Batch matrix-matrix and Dynamic Attention BMM applied a legacy Attention
processor to pipe.unet, which a diffusion transformer does not have, so
they served unet models alone and said nothing elsewhere. The choices, the
processor and its slice helper are removed, an unrecognized method now
warns rather than selecting nothing, and a stored value is rewritten to
Scaled-Dot-Product on load.
attention_slicing holds one of Default, Enabled or Disabled, so testing the
string for truth sent Disabled down the enable branch and left the disable
call unreachable, while the log line below it reported the choice rather
than the action taken.
bypass_sdpa_hijacks and llm_context restore the pinned original over the
router and put the router back, including when the body raises. Captioners
and detailers run inside them, so the router must be removable.
xformers_options had no reader, and Sub-quadratic has not been an
attention choice for a long time, so the hypertile branches keyed on it
never ran. Configs that still store xformers_options load without the
unknown-setting warning.
The sdnq backend read six settings on every call; it now captures them
when the chain is built. Each backend declares the settings its call
captures, and webui registers one onchange over those names plus the
override set and the torch kernel flags, so a change rebuilds the chain
between jobs. When a compiled model is resident the rebuild also resets
dynamo, since its graphs hold the previous router.
SD_ATTN_DEBUG logs each distinct route once: backend, component role,
step, shapes, dtype and mask presence. The router takes an optional
observer for it, so the clean path carries one pointer check. report()
returns the active chain and generation context, and torch_info records
the whole chain as one string instead of the last prepared backend.
A module-level context tells attention consumers what is running: the
component role (transformer, text encoder, vae), the index of the
denoiser forward about to run, the pass length, and the model. It is
opened and closed around process_images, reset per denoising pass beside
the callback setup, and advanced by both step sources: the classic
callback passes the completed step plus one, the modular pre-forward
hook counts forwards. Roles come from the existing text encoder and vae
hijacks and the modular phase hooks. The step also lives in a device
scalar updated in place, so a compiled reader keeps its graph across
steps.
The generate-time gate compared the stored processor name against the
sdp_overrides list, which can never be equal, so the check reduced to
the processor name alone and a changed override set was never applied
until the next model load. set_diffusers_attention now stamps the
override set it applied beside the processor name, the gate compares
both, and pipe switches carry the new attribute with the old one.
The flex backend never called the sdpa it replaced, so any backend
stacked before it was unreachable and every call it could not serve,
cpu or 3d inputs included, failed inside flex_attention. It is now an
ordinary entry gated on what flex_attention accepts: 4d tensors on one
non-cpu device. The mask path drops the 2d special case, which indexed
attn_mask.size and reshaped the mask onto the wrong axis; expanding to
(batch, heads, q, kv) already follows sdpa broadcast semantics.
Replace the six closure hijacks stacked in devices.set_sdpa_params with
a registry of declarative backends and one router installed in their
place. Each backend declares the constraints its closure carried as a
predicate, a priority matching its old stacking position, and a prepare
step that imports and configures the implementation; the router walks
the prepared entries by priority and hands declined calls to the
terminal backend (dynamic, flex) or the original sdpa, so fallback is
the router's job rather than each closure's.
- parity held: gates transcribed literally, the same kernel kwargs,
enable_gqa passed to the original only when set, torch_info keeps the
last prepared backend, the dynamic pin still set
- a backend enabled on a platform without it warns instead of silently
doing nothing
- the legacy set_* entry points are gone; devices.py installs the router
- test/test-attention-router.py checks every override subset against the
old stacking order, gate parity over 16,000 shape cases, dispatch,
terminal handoff and prepare isolation, offline
modules/attention.py becomes modules/attention/: hijacks.py keeps the six
sdpa monkeypatch setters, dispatcher.py the diffusers-side processor and
dispatcher setup with the kernels hub hijack, and the package facade
re-exports every public name so call sites are unchanged. The devices
import moves inside set_diffusers_attention, which removes the
devices <-> attention import cycle.
A delta that does not fit its target module cannot apply, and applying only
the layers that do fit leaves the model in a state nothing was trained for,
so try_load_chain drops the whole file when any family reports a mismatch.
Bias deltas were never checked against the target bias and could only surface
at apply time; a module with no bias stays a non-mismatch, since whole
architectures are built bias=False.
- check bias deltas against the module bias in the lora, norm and full loaders
- carry the mismatch count on the network so the chain can refuse the file
- record refused writes in the infotext so a partial apply is not read as clean
- point the krea2 full-diff test at a module that has a bias
network_add_weights defaulted its base tensor to self.weight for the bias
delta as well, so in fuse mode a diff_b was added to the weight matrix and
the result written into the bias. Layers where in and out differ threw a
shape error and had the weight matrix installed as their bias, square layers
broadcast silently, and either way the summary still counted the delta as
applied.
- pick the base tensor from the bias flag
- name the layer, target and both shapes in the mismatch error
- return which of (weight, bias) took a write, count the rest as refused
- report refused= on partially applied and partially removed networks
- cover both apply paths in test/test-lora-apply.py
CalamitousFelicitousness
Vladimir Mandic
Dity0